Skill Builder for Openclaw

A precision tool for deciding when to create AI skills and building minimal, high-performance skill skeletons.

zurbrick
v1.0.4
Mar 20, 2026
0
1k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install lean-skill-builder

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install lean-skill-builder using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Skill Builder?

Skill Builder is a specialized utility designed for the Openclaw Skills framework to ensure AI agents remain lean and efficient. Instead of acting as a generic factory, it serves as a technical architect that evaluates whether a task truly requires a dedicated skill or if it can be handled via plain file edits, existing tools, or subagents.

By focusing on the philosophy of the smallest useful answer, Skill Builder helps developers avoid bloat and context drift. It provides a structured approach to drafting SKILL.md files, managing reference directories, and auditing existing skills to ensure they remain sharp and trigger-accurate.

Skill Builder Use Cases

  • Deciding if a repetitive task warrants a new entry in your Openclaw Skills library.
  • Drafting minimal SKILL.md skeletons with precise trigger descriptions.
  • Auditing bloated skills to move deep technical details into reference files.
  • Tightening the scope of an agent to prevent accidental activations during unrelated tasks.
  • Restructuring existing skills to align with modern writing patterns and progressive disclosure.

How Skill Builder Works

  1. Run the decision tree to choose between a plain edit, existing tool, subagent, or new skill.
  2. Recommend the most honest, minimal path forward to avoid unnecessary complexity.
  3. If a skill is justified, generate a lean folder structure including the SKILL.md and essential reference stubs.
  4. Apply high-quality writing patterns to the skill body to ensure clear instructions and output templates.
  5. Conduct a smoke test using realistic prompts to verify trigger specificity and edge-case handling.
  6. Perform a final audit using a checklist to remove any stale claims or duplicate guidance.

Skill Builder Setup

To start building lean Openclaw Skills, ensure your environment is configured to support the standard skill directory structure:

# Create a new skill directory
mkdir -p my-new-skill/references

# Initialize the core skill file
touch my-new-skill/SKILL.md

# Audit an existing skill for bloat
openclaw audit ./path/to/existing-skill

Skill Builder Data Schema & Taxonomy

The skill follows a strict hierarchical taxonomy to minimize context window usage for the AI agent:

Component Purpose Requirement
SKILL.md Core logic, scope, and high-level workflow. Mandatory
references/ Technical specs, checklists, and optional deep-dive details. Optional (1-3 files)
assets/ Static templates or bundled files required for output. Optional
scripts/ Deterministic code for repeatable, non-probabilistic execution. Optional
frontmatter YAML metadata including name and description for triggering. Mandatory

Skill Builder Advanced Features

  • Decision Tree Integration: Uses specialized logic to prevent skill proliferation and maintain repository health.
  • Progressive Disclosure Model: Keeps the main SKILL.md thin by delegating complex metadata to the references folder.
  • Smoke Testing Protocol: Built-in validation for clear triggers, edge cases, and near-miss scenarios.
  • Tool Addition Gate: Prioritizes existing primitives and retrieval patterns before suggesting new tool creation.
  • Multi-Agent Handoff: Seamlessly transitions to skill-creator for advanced benchmarking and iteration loops.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*